[0:00:00] Sam, great to have you back on Bloomberg. Always a pleasure. It's been a while for first interview in a very long time. So I want to start with OpenAI because OpenAI was founded in part out of a fear that artificial general intelligence or sort of superhuman intelligence shouldn't be in the hands of any one company and was perhaps coming faster than we think. Is it really just right around the corner or is it farther off than perhaps even you thought it was initially?
[0:00:27] I think it's a question of timeframes. Like I view the next few decades for this sort of most important technological milestone in human history, I view that as right around the corner. And so the debate of is it 10 years or 100 years, I don't think that matters too much given the magnitude of what's happening. Like it's coming soon enough and it's a big enough deal that I think we need to think right now about how we want this deployed, how everyone gets a benefit from it, how we're going to govern it, how we're going to make it safe and sort of good for humanity. Do you think that AI will share our values?
[0:00:56] We have a team working on that right now. So how we impart human values, which are difficult to encode in math, right? So how do we teach a system our collective human values? And I will say I'm more optimistic now than I was before that we will be able to accomplish that goal. Well, and do we even want AI to share our values? Because there are many inequities that exist in society that we may not want to be replicated in AI. One of my hopes for AI is that it will take
[0:01:27] Elon Musk, who founded this with you, has been concerned about the sort of apocalyptic
[0:01:57] Possible future of AI. How realistic is that apocalypse? I think it's always hard to say, like, when you have any incredibly powerful new technology, here's exactly how it's going to go. I will say that I am personally optimistic we are going to get to the good future, but I think that's going to require incredibly hard work from very talented people that needs to start now. So how big a risk is it that that work doesn't happen, or that AI gets off track, or that AI supersedes human intelligence and then has a mind of its own?
[0:02:26] One thing that a lot of people worry about is that, you know, some kid in a garage is going to invent AI by him or herself and then misuse it or not think through all the safety issues. And one thing that I've become more confident about is that that's not going to happen. And the reason is that this requires truly massive compute resources, you know, giant data centers, custom-built chips. And so I think there will be, we'll know all the projects in the world working on it. Some will be government, some will be private, some will be like us.
[0:02:53] But my hope is that as we get closer and closer, those projects will come together and make sure that we're thinking about the safety and the control issues. In looking towards the future, there's some concern that artificial intelligence research and development has hit a wall and that we need a radically different approach to achieve the breakthroughs that we were hoping for. Would you agree with that? Not at all. I think if you look at the landmark result in the field each year for the past, say, four years,
[0:03:22] It's incredibly more impressive every year. You know, it wasn't that long ago we couldn't even beat Go with AI. And now we can beat incredibly sophisticated video games with hundreds of thousands of actions taken with thousands of options at each step. And we have made algorithmic progress each year. We've also made incredible hardware progress. You know, every year we can use 10 times more compute power per deep learning model. The rate of progress is incredible, and that's going to continue. So you recently had a team of bots. Yes.
[0:03:52] Play against some of the best human players in Dota 2. But the humans still won. Is that a setback? No, I don't believe they'll win for much longer. This is how it works. Every month we get better and better. And that's just the rate of... I think basically every two weeks, our bot has like a 90% win rate against the bot from two weeks ago. So the rate of improvement there is just unbelievable. So what's underhyped in the discussion about AI?
[0:04:23] Well, I think people talk about sort of self-driving trucks and the impact that's going to have on jobs, and that's true. But what they don't talk about is all of the white-collar work that AI is going to do. My belief is that all repetitive human work that doesn't require the deep emotional connection between two people, that will all be done in the next couple of decades better, cheaper, faster by AI. And that is somehow left out of the conversation. It's always about factory robots and self-driving cars. What's overhyped?
[0:04:54] The killer robots thing gets overdone. No robots are going to eat all our jobs or kill us? Some robots will eat some jobs, and some robots will kill humans. You've already seen that with self-driving cars. But when you look at the trend as a whole, I think it's going to be incredibly positive for humanity. But you do believe that AI can supersede human intelligence? I believe it absolutely will. There's a big debate about time frames,
[0:05:23] I think it takes unique human arrogance to believe that AI cannot supersede humans. But what about that scares you? So many things, right? This is like, what does it mean to build something that is more capable than ourselves? What does that say about our humanity? What's that world going to look like? What's our place in that world? How is that going to be equitably shared? How do we make sure that it's not like a handful of people in San Francisco making the decisions and reaping all the benefits? I think we have...
[0:05:52] An opportunity that comes along only every couple of centuries to redo the socioeconomic contract and how we include everybody in that and make everybody a winner and how we don't destroy ourselves in the process is a huge question. You've shared some controversial thoughts about universal basic income. How is the work you're seeing in AI impacting your view of social inequality? Well, one is that I think
[0:06:17] The social inequality we have today is nothing compared to what we could have if we don't address it because AI will just be such a powerful lever. Two is that I think we do have an opportunity as we rewrite the social contract to think about how we can get towards a more equal world. We'll never get all the way there, of course, and human flaws run deep. But I think we're going to have an opportunity to push the reset button and think about the world we want. And I think universal basic income is one part of that.
[0:06:46] You know, how we give like, I don't know what to call it, universal basic meaning. We'll have plenty of resources for everyone. But then how do we make everyone happy and fulfilled and find their purpose? You know, we get to find out. You're still spending half your time at Y Combinator. What trends are you seeing when it comes to, you know, the value of startups that are being created, the valuation of the startups that are being created, and whether it's worth it?
[0:07:15] You know, I made this public bet three years ago, four years ago, something like that, when everyone was saying there was a bubble. And I just looked at it recently. It's looking pretty good. And at the time, everyone was like, this is ridiculous. You're going to lose the bet by far. And everyone was saying there was a bubble, so I suspected there wasn't. Now, most people aren't saying there's a bubble, and so I'm, of course, nervous. Everyone's now back to operating mode where everything's great, everything's going to keep going up.
[0:07:43] So I'm much more nervous than I was when I made that post. I think it was like three and a half years ago. I try not to get too focused on bubbles because I try to keep this multi-decade view. And I want to invest in shares that we're going to hold through multiple busts and multiple bubbles over the decades of compounding it takes to create a super valuable company. However, when you look around at the moment, sure, plenty of stuff seems out of whack. But on the time frame...
[0:08:12] I don't care that much. In terms of trends, one trend that we're seeing that's affecting all startups now, at least in the Bay Area, is the competition for talent with the tech hypercaps is very difficult. When you have Google or Facebook finding the best engineers and putting these incredibly large compensation packages in front of them, how you as a startup are...
[0:08:41] Attract enough talent, concentrate enough talent in your team to win. It's become a very hard question. In fact, that's become one of the first questions I ask startups. Like, what's your approach to this? Are you going to hire less people? Are you going to give them a huge amount of equity? Is the mission so important they're going to join anyway? And it is astonishing how many people don't have an answer to that question. So how do they compete? Oh, it's always different. Like, any of those things I just said are one approach. But the problem is that most people seem to have not thought about it at all.
[0:09:08] Just to put a fine point on it, you try not to think about whether or not we're in a bubble, but you think we could be in a bubble right now. I mean, I think we are. I think some valuations are too high. I think some are too low. I think that if interest rates go up a lot, I think that probably has some ripple effect, I hope, on equity valuations. But I think it's not that binary of a question. It feels more likely to be true than it was in 2015.
[0:09:37] But again, there's still a lot of startups that I put a lot of money into at current valuation, so I'm not smart enough to figure that out. Speaking of values, you recently suspended your role on the board of the Saudi Future City. Why did you decide to do that? I don't have a lot more about it to say than I did before, but I feel way out of my area of expertise from what I can see. I have a lot of questions, and I'd like to wait for those to be answered.
[0:10:04] Hopefully very soon before deciding about my involvement in the board. In general, I believe you should always wait for the data to make any sort of decision, any sort of comment. But this case seemed to be exceptional. We seem to be entering a prolonged period of tech hating after coming out of a period of tech adulation, praise. And there's a lot of angst and anger towards Facebook and Google about privacy, about trust. Does that last? And does that...
[0:10:32] I don't know. I feel this is also well outside of my area of expertise. I generally believe you should believe people when they're upset, but not always about why. Because I think it's hard to articulate exactly why aren't people mad at the tech companies. And I have my own reasons, things I'm upset about, and I'm sure you have your own. And I'm sure they're very different from other people that live in different contexts than we do.
[0:11:02] I think fundamentally these companies still do add a lot of value to a lot of people's lives and people love that. And I think they're being pretty responsive to changes. And I think, you know, like everyone wants to do better. I fundamentally believe that. I don't believe the trope that these are like companies that are solely motivated by profit and they don't care about their users and they don't care about doing good for the world. I think they fundamentally do. And I think they're trying to adapt and I think they will.
[0:11:29] I think it's like a deeply American value to be suspicious of huge concentrated power and wealth. And so I think we're seeing that. And I think that's good. You see the next generation of companies that are trying to disrupt Facebook and Google. Do you think they can be disrupted in part as a result of this? Not as a result of this. I just think you can always disrupt. At any given moment in history, the giant companies seem unassailable. And yet...
[0:11:55] There's always a new generation. The next Facebook and Google will not look like Facebook and Google in the same way that Facebook and Google didn't look like Intel and Microsoft. But they will come. And I don't know what they'll be. You don't know what they'll be. But they will be something. And they will be huge. And they'll be new. And they'll be great. And that's what drives our system forward. And that's why our quality of life gets better every decade. I will say one thing about the new generation of companies that I've noticed is that they are more thoughtful at the beginning about the impact that
[0:12:24] Their companies will have, if they're wildly successful, on the world and people's lives than I think the companies from 10 or 15 years ago were. And I think that's part of growing up in this kind of environment. I'm happy about that.